[Mlir-commits] [mlir] Add Vector bitwidth target to Linearize Vectorizable and Constant Ops (PR #83314)

[Ivan Butygin]

================
@@ -19,10 +19,27 @@
 
 using namespace mlir;
 
+static bool isLessThanTargetBitWidth(Operation *op, unsigned targetBitWidth) {
+  auto resultTypes = op->getResultTypes();
+  for (auto resType : resultTypes) {
+    VectorType vecType = cast<VectorType>(resType);
+    unsigned trailingVecDimBitWidth =
+        vecType.getShape().back() * vecType.getElementTypeBitWidth();
----------------
Hardcode84 wrote:

I think, vectors of index type are allowed, and `getElementTypeBitWidth()` will crash for them, we need to handle them somehow (probably just ignoring them entirely in this pass).

https://github.com/llvm/llvm-project/pull/83314